The base installation of R contains several functions for plotting and visualizing data. In addition to these, there are several add-on packages that greatly extend the basic functionality. Here, we will look at one of the most widely used of these, ggplot2. Before starting the lab, you will need to set up a new folder for your working directory. Go to your geog6000 folder now and create a new folder for today’s class called lab05b. Make sure to set this to the working directory with setwd(). As a reminder, you can also do this by going to the [Session] menu in RStudio, then [Change working directory]. This will open a file browser that you can use to browse through your computer and find the folder.

The ggplot2 package was written by Hadley Wickham, and implements the Grammar of Graphics framework developed by Leland Wilkinson. The idea behind the grammar of graphics is that all plots can be described by a common language, rather than considering them as separate barplots, line charts, etc. What differs is the coordinate system and geometry used to place the data on the page. Using this package requires a little more work than standard plots, but the results are usually worth while.

The ggplot2 package comes with two main functions: qplot(), a simple plotting routine which is designed to replace the base plot() function in R; and ggplot(), which provides a much greater range of functions. We’ll look briefly at the first of these before looking at ggplot() in more detail.

Start by installing and loading the ggplot2 package:

install.packages("ggplot2")
library(ggplot2)

We’ll use several datasets in this lab:

  • A dataset of morphological measurements of three species of penguin penguins.csv
  • a dataset on the growth of five Orange trees (orange.csv)
  • Gapminder data on countries population, GDP and life expectancy for the past 60 years (gapminderData.csv)
  • The VADeaths dataset (VADeaths.csv), showing death rates in West Virginia

You will need to download these files from Canvas, and move them from your Downloads folder to the datafiles folder that you made previously. Then load these data sets and use str() and names() to look at what they contain:

penguins <- read.csv("../datafiles/penguins.csv")
str(iris)
## 'data.frame':    150 obs. of  5 variables:
##  $ Sepal.Length: num  5.1 4.9 4.7 4.6 5 5.4 4.6 5 4.4 4.9 ...
##  $ Sepal.Width : num  3.5 3 3.2 3.1 3.6 3.9 3.4 3.4 2.9 3.1 ...
##  $ Petal.Length: num  1.4 1.4 1.3 1.5 1.4 1.7 1.4 1.5 1.4 1.5 ...
##  $ Petal.Width : num  0.2 0.2 0.2 0.2 0.2 0.4 0.3 0.2 0.2 0.1 ...
##  $ Species     : Factor w/ 3 levels "setosa","versicolor",..: 1 1 1 1 1 1 1 1 1 1 ...
orange <- read.csv("../datafiles/orange.csv")
str(orange)
## 'data.frame':    35 obs. of  3 variables:
##  $ Tree         : int  1 1 1 1 1 1 1 2 2 2 ...
##  $ age          : int  118 484 664 1004 1231 1372 1582 118 484 664 ...
##  $ circumference: int  30 58 87 115 120 142 145 33 69 111 ...
gapdata <- read.csv("../datafiles/gapminderData5.csv")
str(gapdata)
## 'data.frame':    1704 obs. of  6 variables:
##  $ country  : chr  "Afghanistan" "Afghanistan" "Afghanistan" "Afghanistan" ...
##  $ year     : int  1952 1957 1962 1967 1972 1977 1982 1987 1992 1997 ...
##  $ pop      : num  8425333 9240934 10267083 11537966 13079460 ...
##  $ continent: chr  "Asia" "Asia" "Asia" "Asia" ...
##  $ lifeExp  : num  28.8 30.3 32 34 36.1 ...
##  $ gdpPercap: num  779 821 853 836 740 ...
VADeaths <- read.csv("../datafiles/VADeaths.csv")
str(VADeaths)
## 'data.frame':    5 obs. of  5 variables:
##  $ Age         : chr  "50-54" "55-59" "60-64" "65-69" ...
##  $ Rural.Male  : num  11.7 18.1 26.9 41 66
##  $ Rural.Female: num  8.7 11.7 20.3 30.9 54.3
##  $ Urban.Male  : num  15.4 24.3 37 54.6 71.1
##  $ Urban.Female: num  8.4 13.6 19.3 35.1 50

Reshaping tables for plotting

ggplot2 is part of an opinionated ecosystem of R packages called the Tidyverse. Without going too far into the weeds, one of its core tenets is that data should be in long format, so all of its R packages just assume that the data have this form, including ggplot2. What does this mean exactly? Well, have a look at the VADeaths table again. It is currently in a wide format:

VADeaths
##     Age Rural.Male Rural.Female Urban.Male Urban.Female
## 1 50-54       11.7          8.7       15.4          8.4
## 2 55-59       18.1         11.7       24.3         13.6
## 3 60-64       26.9         20.3       37.0         19.3
## 4 65-69       41.0         30.9       54.6         35.1
## 5 70-74       66.0         54.3       71.1         50.0

Here, we have a categorical column for each demographic class: (i) rural and male, (ii) rural and female, (iii) urban and male, and (iv) urban and female. And the values in each cell represent the death rate for that demographic class and age range. To make this table long form (to reshape it), we would want four columns: Age, Residence, Gender, and Death Rate; with the Residence and Gender columns containing the data held unhelpfully in the column names.

One very useful R package for reshaping tables in this way is tidyr, whose primary functions are pivot_wider and pivot_longer, which reshape the data into wide and long formats, respectively.

library(tidyr)

VADeaths2 <- pivot_longer(VADeaths, 
                          cols = Rural.Male:Urban.Female,
                          names_to = c("Residence", "Gender"), 
                          names_sep = "\\.",                   
                          values_to = "DeathRate")             

VADeaths2
## # A tibble: 20 x 4
##    Age   Residence Gender DeathRate
##    <chr> <chr>     <chr>      <dbl>
##  1 50-54 Rural     Male        11.7
##  2 50-54 Rural     Female       8.7
##  3 50-54 Urban     Male        15.4
##  4 50-54 Urban     Female       8.4
##  5 55-59 Rural     Male        18.1
##  6 55-59 Rural     Female      11.7
##  7 55-59 Urban     Male        24.3
##  8 55-59 Urban     Female      13.6
##  9 60-64 Rural     Male        26.9
## 10 60-64 Rural     Female      20.3
## 11 60-64 Urban     Male        37  
## 12 60-64 Urban     Female      19.3
## 13 65-69 Rural     Male        41  
## 14 65-69 Rural     Female      30.9
## 15 65-69 Urban     Male        54.6
## 16 65-69 Urban     Female      35.1
## 17 70-74 Rural     Male        66  
## 18 70-74 Rural     Female      54.3
## 19 70-74 Urban     Male        71.1
## 20 70-74 Urban     Female      50

There are three important arguments to pivot_longer in addition to stating what columns to pivot. First, names_to specifies the names of new columns for demographic class data. Second, names_sep specifies the character for separating the column names as values into the respective columns; in this case, the period (‘.’), which is a reserved expression and thus needs to be escaped with two back ticks (‘\’). Finally, values_to specifies the name of the new column for death rate data.

Simple plots with qplot()

Scatterplots

A basic scatterplot. Note that the syntax matches the original plot() function, with the \(x\) variable specified first, and the \(y\) variable second, but the output is quite different:

qplot(bill_depth_mm, 
      bill_length_mm, 
      data = penguins)

Coloring the symbols by species (note that a legend is automatically added):

qplot(bill_depth_mm, 
      bill_length_mm, 
      data = penguins,
      color = species)

Adding further information by resizing the symbols with a third variable (petal width):

qplot(bill_depth_mm, 
      bill_length_mm, 
      data = penguins,
      color = species, 
      size = body_mass_g)

Finally, we add labels and a title, and make the symbols transparent using the parameter alpha, in order to make it clear where these overlap:

qplot(bill_depth_mm, 
      bill_length_mm, 
      data = penguins,
      color = species, 
      size = body_mass_g, 
      alpha = I(0.7),
      xlab = "Bill depth (mm)", 
      ylab = "Bill length (mm)",
      main = "Palmer Penguin Data")

By default, qplot() makes a scatterplot, but this can be changed to a different plot type or geometry, by using the geom argument. For example, to make a histogram:

qplot(bill_depth_mm, 
      data = penguins, 
      geom = 'histogram', 
      binwidth = 0.25)

Or a set of boxplots:

qplot(species, 
      bill_depth_mm, 
      data = penguins, 
      geom = 'boxplot')

Line plots can be made with geom = "line". Try this with the set of Orange trees:

qplot(age, 
      circumference, 
      data = orange, 
      geom = 'line')

As this is plotting the circumference from all five trees in the dataset together, the line doesn’t make much sense. We can separate out the different trees by telling qplot to use the tree number as a color. As this is a continuous variable, we first convert this into a factor to indicate the different groups (trees):

orange$Tree = factor(orange$Tree)
qplot(age, 
      circumference, 
      data = orange, 
      color = Tree, 
      geom = 'line')

More advanced figures with ggplot()

The ggplot() function offers more control over your plots. It works in quite a different way to the other plotting functions, that start with a base plot, then add other points, lines, etc to the figure. Instead, ggplot creates a plot object, which can be adjusted and added to as you proceed.

In order to understand how ggplot makes a figure, we need to establish what the fundamental parts are of every data graph. They are:

  1. Aesthetics – these are the roles that the variables play in each graph. A variable may control where points appear, the color or shape of a point, the height of a bar and so on.
  2. Geometries – these are the geometric objects which represent the data: points, lines, bars, …
  3. Statistics – these are the functions which add some interpretation to the data, e.g. best fit line, location of median, etc
  4. Scales – these are legends that show the relationship between variables and different symbols or colors (e.g. circular symbols represent females while squares represent males)
  5. Facets – these are groups in your data which may be used to make multiple graphs, each for one of the groups. For example, faceting by gender would cause the graph to repeat for the two genders.

When making a ggplot figure, we generally start by creating the base figure. To do this we need to tell the function where the data is coming from, and the base aesthetic (i.e. which variable is x, which is y?). To remake the first iris plot with ggplot(), we would do the following:

myplot <- ggplot(penguins, aes(x = bill_depth_mm, y = bill_length_mm))

If you now type myplot at the command line, R will complain that there are no layers in the plot. This is because we need to tell ggplot what geometry to use. To do this, we simply ‘add’ one of ggplot’s geometry functions. In this case, we use geom_point(), to make a scatterplot:

myplot + geom_point()

A very important thing to note is that this has not changed our base object myplot, but simply displayed it with the point geometry added to it. If you retype myplot at the command line, you will again be told that there are no layers. In order to make this change stick (i.e. keep the use of the point geometry), we need to resave the plot object:

myplot <- myplot + geom_point()

Now typing myplot makes the plot appear. This allows you to experiment with different displays, and only update (or resave) the plot object when you have changes that you want to keep.

We could now add other geometries, statistics or other effects to the plot figure to change the way that it looks. Instead, we’ll work through several different examples now, using the Gapminder dataset to look at the different functions that are available.

Histograms

A simple histogram of life expectancy for all countries:

myplot <- ggplot(gapdata, aes(x = lifeExp))

myplot + geom_histogram()

Let’s add a title and axis labels to the base plot and remake the histogram. Note that by adding these to the base plot object, all subsequent plots will have these labels and title:

myplot <- myplot + labs(x = "Life Expectancy", title = "GapMinder Life Expectancy")

myplot + geom_histogram()

The output should tell you that the histogram binwidth is set to 1/30 of the data range. This can easily be adjusted:

myplot + geom_histogram(binwidth = 1)

We might be interested in looking at the histogram of values by continent. There are a couple of ways to do this. The first is to include an aesthetic which links the continent to the color used to fill the histograms. Note that we do this by adding the aesthetic to the geom_* function (you could equally remake the original ggplot() command with this included):

myplot + geom_histogram(aes(fill = continent))

By default, ggplot stacks the histograms - a better way to do this is have all bars starting at zero and superimposed. This can be achieved with the position parameter:

myplot + geom_histogram(aes(fill = continent), position = "identity")

An alternative display with several groups is to use facets. There are two types of facets in ggplot2: facet_wrap which splits plots based on one group, and facet_grid which splits plots based on two groups. We’ll start by using facet_wrap to display the histograms by continent. We also add two parameters to geom_histogram to change the border and fill color of the histograms:

myplot + 
  geom_histogram(color = 'lightblue', fill = 'white') + 
  facet_wrap(~ continent)

The dataset also includes the year that the data was taken in. We can use this as a second grouping, which allows us to make histograms per continent for every year that data was recorded:

myplot + 
  geom_histogram(fill = 'darkorange') + 
  facet_grid(year ~ continent)

Density plots

An alternative display of univariate, continuous data is through density plots (geom_density()). These fit a small Gaussian window around each observation, then add the value of all the windows across the range of data. The result is a smoothed, continuous looking histogram. As before, we can use the fill parameter to separate out continents. We add an alpha value again to make the results transparent:

myplot + geom_density(aes(fill = continent), alpha = 0.4)

Boxplots

We have already looked at boxplots using the base R functions. Here, we need to specify the grouping within the plot aesthetic as x, and the continuous variable as y. As we need both x and y, we need to make a new ggplot object:

myplot <- ggplot(gapdata, aes(x = continent, y = lifeExp))

myplot + geom_boxplot()

By default, the order of the groups is alphabetical, but it might be useful to order the boxplots to reflect something about the data, for example from low to high average life expectancy. The R function reorder can be used here. This require three parameters: a factor to be reorganized, a variable to be used to reorganize it, and a function to be calculated on that variable. Here we reorder by median life expectancy:

myplot + geom_boxplot(aes(x = reorder(continent, lifeExp, median)))

Barcharts

Two types of barchart can be produced by ggplot2. By default, the barcharts are presented as summary of the variable of interest. With factor or categorical variables, this is simply a count of each category. For example:

myplot <- ggplot(gapdata, aes(x = continent))

myplot + geom_bar()

Often, however, we want to use some precalculated value for the height of the bar. In the Virginia death rate dataset, for example, the death rates have already been calculated by age and demographic class. To plot death rates per age groups, we can do the following:

myplot <- ggplot(VADeaths2, aes(x = Age, y = DeathRate))

myplot + geom_bar(stat = "identity")

We can split the plot into the different demographic groups. As before, we can do this in two ways. First, by including the demographic group as a fill in the aesthetic setup:

myplot + geom_bar(aes(fill = Residence), stat = "identity")

myplot + geom_bar(aes(fill = Residence), stat = "identity", position = "dodge")

Second, by using faceting the data using facet_wrap() (we also use coord_flip() to make the bars horizontal):

myplot + 
  geom_bar(aes(fill = Residence), stat = "identity", position = "dodge") + 
  facet_wrap(~ Gender) + 
  coord_flip()

Scatterplots

Scatterplots are the simplest way to display bivariate data. ggplot2 has several different geometries, including points and lines. To make a simple figure of life expectancy against per capita GDP from the gapminder dataset:

myplot <- ggplot(gapdata, aes(x = gdpPercap, y = lifeExp)) 

myplot + geom_point()

Given the distribution of the GDP variable, a clearer display could be obtained by transforming it into log10 values. We can do this simply by adding the transformation in call to ggplot():

ggplot(gapdata, aes(x = log10(gdpPercap), y = lifeExp)) + geom_point()

But we can also manipulate the axis scale directly, which keeps the original values on the axis labels:

myplot + 
  geom_point() + 
  scale_x_log10()

As this will be useful for future plots, will modify the underlying ggplot object:

myplot <- myplot + scale_x_log10()

A further type of geometry (geom_smooth()) allows us to add trend lines to the plot.

myplot + 
  geom_point() + 
  geom_smooth()

The basic method uses either a type of spline or a local regression (loess) to fit the line. A simple linear fit can be added using the parameter method='lm'

As before, we can use an extra variable to identify different parts of the data set. The Gapminder data contains a column indicating which continent the observations came from. We’ll use this first to color the points by changing the aesthetic:

myplot + geom_point(aes(color = continent))

If we now add a best fit line, however, we only get a single line for the whole dataset:

myplot + 
  geom_point(aes(color = continent)) + 
  geom_smooth(method = 'lm')

This is a little tricky to explain, but basically, we have only changed the aesthetic in the geometry of the points, rather than the underlying ggplot object. The simplest way around this is to remake the object with the color included, then add the smoother:

myplot <- ggplot(gapdata, aes(x = gdpPercap, y = lifeExp, color = continent)) + scale_x_log10()

myplot + 
  geom_point() + 
  geom_smooth(method='lm')

Note that if you drop the geom_point(), you will just obtain the trend lines.

We can also split the data into facets, which will automatically split the data when adding trend lines:

myplot <- ggplot(gapdata, aes(x = gdpPercap, y = lifeExp)) + scale_x_log10()

myplot + 
  geom_point() + 
  facet_wrap(~ continent)

And we can add a linear fit by continent:

myplot + 
  geom_point() + 
  facet_wrap(~ continent) + 
  geom_smooth(method = 'lm')

Finally, we can facet by both continent and year by using facet_grid():

myplot + 
  geom_point() + 
  facet_grid(year ~ continent) + 
  geom_smooth(method = 'lm')

Line plots are a good method for displaying data over time. We’ll make a spaghetti diagram of the change in life expectancy for each country, broken out by continent. To do this, we tell the geom_line() function to use groupings based on the country name for each line:

myplot <- ggplot(gapdata, aes(x = year, y = lifeExp))

myplot + 
  facet_wrap(~ continent) + 
  geom_line(aes(group = country, color = continent))

We can also look at different subsets of the data using a couple of different methods. Both of these examples used the R function subset() to extract just part of a dataframe; the first just extracts a single country, the second all countries from a list:

ggplot(subset(gapdata, country == "Zimbabwe"),
       aes(x = year, y = lifeExp)) + 
  geom_line() + 
  geom_point()

jCountries <- c("Canada", "Rwanda", "Cambodia", "Mexico")

ggplot(subset(gapdata, country %in% jCountries),
       aes(x = year, y = lifeExp, color = country)) + 
  geom_line() + 
  geom_point()

Saving your ggplot2 figures

In this section we’ll make a figure and save it to a file. We’ll start by building the figure, then add various elements, including axis labels and a title, and finally save it to a file using ggsave(). Note that as we make the figure, we update the `ggplot`` object each time:

Plot life expectancy against GDP with colors by continent

myplot <- ggplot(gapdata, aes(x = gdpPercap, y = lifeExp, color = continent)) + scale_x_log10()

myplot <- myplot + 
  geom_point() + 
  geom_smooth(method = 'lm')

Now add a title and labels:

myplot <- myplot + 
  ggtitle("Life Expectancy vs. GDP") + 
  xlab("GDP") + 
  ylab("Life Expectancy")

myplot

To save the image, you can dump it directly from the R plot window, but a better way is to use the ggsave() function. This provides more control over the final figure. To save as a png:

ggsave("lifeexp_vs_gdp.png", myplot)

As a pdf (jpeg, svg, eps also available):

ggsave("lifeexp_vs_gdp.pdf", myplot)

To adjust the output size:

ggsave("lifeexp_vs_gdp.png", 
       myplot,
       width = 6, 
       height = 4)

Themes

ggplot2 has a default theme with a gray background to the plot. There are a number of included themes that can simply applied to change this. Two of the built in themes are ‘classic’ and ‘bw’:

myplot + theme_classic()

myplot + theme_bw()

The add-on package ggthemes includes a number of different themes, including themes inspired by the Economist, the Wall Street Journal and Nate Silver’s blog 538. You will need to install and load this package, for this to work.

library(ggthemes)

myplot + theme_calc()

myplot + theme_economist()

myplot + theme_wsj()

myplot + theme_fivethirtyeight()